Download Imaging transport: optical measurements of

Document related concepts

Surface plasmon resonance microscopy wikipedia , lookup

Auger electron spectroscopy wikipedia , lookup

Phase-contrast X-ray imaging wikipedia , lookup

Magnetic circular dichroism wikipedia , lookup

Reflection high-energy electron diffraction wikipedia , lookup

Fluorescence correlation spectroscopy wikipedia , lookup

Diffraction topography wikipedia , lookup

Microscopy wikipedia , lookup

Ellipsometry wikipedia , lookup

Optical coherence tomography wikipedia , lookup

Scanning joule expansion microscopy wikipedia , lookup

X-ray fluorescence wikipedia , lookup

Ultrafast laser spectroscopy wikipedia , lookup

Scanning tunneling spectroscopy wikipedia , lookup

Scanning electrochemical microscopy wikipedia , lookup

Rutherford backscattering spectrometry wikipedia , lookup

Gaseous detection device wikipedia , lookup

Ultraviolet–visible spectroscopy wikipedia , lookup

Scanning electron microscope wikipedia , lookup

Vibrational analysis with scanning probe microscopy wikipedia , lookup

Photon scanning microscopy wikipedia , lookup

Chemical imaging wikipedia , lookup

Transcript
Calhoun: The NPS Institutional Archive
Theses and Dissertations
Thesis and Dissertation Collection
2004-09
Imaging transport: optical measurements of
diffusion and drift in semiconductor materials and devices
Freeman, Will
Monterey California. Naval Postgraduate School
http://hdl.handle.net/10945/1428
NAVAL
POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA
THESIS
IMAGING TRANSPORT: OPTICAL
MEASUREMENTS OF DIFFUSION AND DRIFT IN
SEMICONDUCTOR MATERIALS AND DEVICES
by
Will Freeman
September 2004
Thesis Advisor:
Nancy M. Haegel
Approved for public release; distribution is unlimited
THIS PAGE INTENTIONALLY LEFT BLANK
REPORT DOCUMENTATION PAGE
Form Approved OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including
the time for reviewing instruction, searching existing data sources, gathering and maintaining the data needed, and
completing and reviewing the collection of information. Send comments regarding this burden estimate or any
other aspect of this collection of information, including suggestions for reducing this burden, to Washington
headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite
1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project
(0704-0188) Washington DC 20503.
1. AGENCY USE ONLY (Leave blank)
2. REPORT DATE
September 2004
3. REPORT TYPE AND DATES COVERED
Master’s Thesis
4. TITLE AND SUBTITLE:
5. FUNDING NUMBERS
Imaging Transport: Optical Measurements of Diffusion and Drift in Semiconductor
Materials and Devices
6. AUTHOR:
Will Freeman
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)
8. PERFORMING ORGANIZATION
Naval Postgraduate School
REPORT NUMBER
Monterey, CA 93943-5000
9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING
AGENCY REPORT NUMBER
11. SUPPLEMENTARY NOTES
The views expressed in this thesis are those of the author and do not reflect the official policy or position of the Department of
Defense or the U.S. Government.
12a. DISTRIBUTION / AVAILABILITY STATEMENT
12b. DISTRIBUTION CODE
Approved for public release; distribution is unlimited
13. ABSTRACT (maximum 200 words)
Knowledge of transport parameters is important to the development of new optoelectronic materials and devices, such as
ultraviolet (UV) semiconductor lasers and advanced solar cells. A series of experiments was performed to measure
fundamental transport parameters in luminescent semiconductor materials. Using a technique that couples a scanning electron
microscope (SEM) in spot mode with a charge coupled display (CCD) camera, it is possible to image the recombination of
charge created at a point. The goal is to extract fundamental transport parameters, such as minority carrier diffusion length
(Ldiffusion)) and drift length (Ldrift), with high spatial resolution. Direct transport imaging was used to study diffusion without bias
and drift under a range of applied electric fields. The recombination distribution as a function of applied bias was imaged. For
the unbiased measurements, the results showed that for bulk n-type GaAs the spotwidth was independent of probe current
indicating the luminescence distribution is primarily a function of generation volume and not diffusion length. For thin layer
samples that could be approximated as two dimensional (2D), it was found that the spotwidth changed as a function of probe
current indicating the potential to extract diffusion length data. Results are compared to numerical modeling of charge transport
and the feasibility and limitations of this method for contact-free measurements of lifetime (τ) and mobility (µ) are assessed.
14. SUBJECT TERMS
Contact-less Measurements, Diffusion, Drift, Semiconductors, Transport Imaging
17. SECURITY
CLASSIFICATION OF
REPORT
Unclassified
18. SECURITY
CLASSIFICATION OF THIS
PAGE
Unclassified
NSN 7540-01-280-5500
15. NUMBER OF
PAGES
61
16. PRICE CODE
20. LIMITATION
OF ABSTRACT
19. SECURITY
CLASSIFICATION OF
ABSTRACT
Unclassified
UL
Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. 239-18
i
THIS PAGE INTENTIONALLY LEFT BLANK
ii
Approved for public release; distribution is unlimited
IMAGING TRANSPORT: OPTICAL MEASUREMENTS OF DIFFUSION AND
DRIFT IN SEMICONDUCTOR MATERIALS AND DEVICES
Will Freeman
DOD Civilian, NAWCWD China Lake, California
B.S., California State University, Northridge, 1996
Submitted in partial fulfillment of the
requirements for the degree of
MASTER OF SCIENCE IN PHYSICS
NAVAL POSTGRADUATE SCHOOL
September 2004
Author:
Will Freeman
Approved by:
Professor Nancy M. Haegel
Thesis Advisor
Professor James H. Luscombe
Second Reader
Professor James H. Luscombe
Chairman, Department of Physics
iii
THIS PAGE INTENTIONALLY LEFT BLANK
iv
ABSTRACT
Knowledge of transport parameters is important to the development of new
optoelectronic materials and devices, such as ultraviolet (UV) semiconductor lasers and
advanced solar cells. A series of experiments was performed to measure fundamental
transport parameters in luminescent semiconductor materials. Using a technique that
couples a scanning electron microscope (SEM) in spot mode with a charge coupled
display (CCD) camera, it is possible to image the recombination of charge created at a
point. The goal is to extract fundamental transport parameters, such as minority carrier
diffusion length (Ldiffusion)) and drift length (Ldrift), with high spatial resolution. Direct
transport imaging was used to study diffusion without bias and drift under a range of
applied electric fields. The recombination distribution as a function of applied bias was
imaged. For the unbiased measurements, the results showed that for bulk n-type GaAs the
spotwidth was independent of probe current indicating the luminescence distribution is
primarily a function of generation volume and not diffusion length. For thin layer
samples that could be approximated as two dimensional (2D), it was found that the
spotwidth changed as a function of probe current indicating the potential to extract
diffusion length data. Results are compared to numerical modeling of charge transport
and the feasibility and limitations of this method for contact-free measurements of
lifetime (τ) and mobility (µ) are assessed.
v
THIS PAGE INTENTIONALLY LEFT BLANK
vi
TABLE OF CONTENTS
I.
INTRODUCTION............................................................................................1
A.
TRANSPORT IN SEMICONDUCTORS..........................................1
B.
PURPOSE OF THESIS.......................................................................1
C.
MILITARY RELEVANCE.................................................................2
D.
THESIS OVERVIEW .........................................................................2
II.
TRANPORT PARAMETERS ........................................................................3
A.
DIFFUSION AND DRIFT ..................................................................3
B.
LUMINECSCENCE IN SEMICONDUCTORS...............................6
III.
IMAGING TRANSPORT .............................................................................11
A.
DIRECT TRANSPORT IMAGING ................................................11
B.
DRIFT MEASUREMENTS..............................................................13
C.
DIFFUSION MEASUREMENTS ....................................................19
1.
Bulk GaAs...............................................................................19
2.
Epitaxial AlGaAs ...................................................................23
3.
GaAs/GaNAs Quantum Well ................................................27
a.
Panchromatic ..............................................................28
b.
950 nm Long Pass Filter, GaNAs Layer
Luminescence..............................................................32
c.
950 nm Short Pass Filter, GaAs Layer
Luminescence..............................................................35
D.
COMPARISON TO COMPUTER MODEL ..................................40
IV.
CONCLUSIONS ............................................................................................43
LIST OF REFERENCES ..........................................................................................45
INITIAL DISTRIBUTION LIST .............................................................................47
vii
THIS PAGE INTENTIONALLY LEFT BLANK
viii
LIST OF FIGURES
Fig. 1.
Fig. 2.
Fig. 3.
Fig. 4.
Fig. 5.
Fig. 6.
Fig. 7.
Fig. 8.
Fig. 9.
Fig. 10.
Fig. 11.
Fig. 12.
Fig. 13.
Fig. 14.
Fig. 15.
Fig. 16.
Fig. 17.
Fig. 18.
Fig. 19.
Fig. 20.
Fig. 21.
Fig. 22.
Fig. 23.
Fig. 24.
Fig. 25.
Excess Minority Carrier Distribution as a Function
of Distance from Generation Point ......................................................10
SEM and CCD Camera........................................................................11
Images of Bulk n-Type GaAs ..............................................................12
Sample 329SI Contacts ........................................................................13
Charge Motion Under Applied Bias ....................................................15
Charge Motion Under Applied Bias, Subtracted .................................17
Bulk GaAs, Luminescent Spot, Raw Data and Spline
Interpolated Data..................................................................................20
Bulk GaAs, Luminescent Spot for Accelerating Voltages
15 to 35 kV...........................................................................................21
Bulk GaAs, Spotwidth as a Function of Accelerating Voltage ...........21
Bulk GaAs, Luminescent Spot for Probe Currents
1x10-10 to 1x10-7 A...............................................................................22
Bulk GaAs, Spotwidth as a Function of Probe Current .......................22
Excitation Volume, 2D and 3D Sample...............................................24
1 µm AlGaAs, Luminescent Spot for Probe Currents
1x10-10 to 1x10-7 A...............................................................................25
1 µm AlGaAs, Spotwidth as a Function of Probe Current ..................25
1 µm AlGaAs and Bulk GaAs, Spotwidth as a Function of
Probe Current .......................................................................................26
GaAs/GaNAs, Panchromatic, Raw Data, Luminescent Spot
for Probe Current 1x10-7 A ..................................................................29
GaAs/GaNAs, Panchromatic, Spline Data, Luminescent Spot
for Probe Current 1x10-7 A ..................................................................29
GaAs/GaNAs, Sample n554a, Panchromatic,
Spotwidth vs Spatial Samples..............................................................30
GaAs/GaNAs, Sample n546, Panchromatic,
Spotwidth vs Spatial Samples..............................................................30
GaAs/GaNAs, Sample n553, Panchromatic,
Spotwidth vs Spatial Samples..............................................................31
GaAs/GaNAs, LPF GaNAs Luminescence, Raw Data,
Luminescent Spot for Probe Current 1x10-7 A ....................................33
GaAs/GaNAs, LPF GaNAs Luminescence, Spline Data,
Luminescent Spot for Probe Current 1x10-7 A ....................................33
GaAs/GaNAs, Sample n554a, LPF GaNAs Luminescence,
Spotwidth vs Spatial Samples..............................................................34
GaAs/GaNAs, Sample n546, LPF GaNAs Luminescence,
Spotwidth vs Spatial Samples..............................................................34
GaAs/GaNAs, Sample n553, LPF GaNAs Luminescence,
Spotwidth vs Spatial Samples..............................................................35
ix
Fig. 26.
Fig. 27.
Fig. 28.
Fig. 29.
Fig. 30.
Fig. 31.
Fig. 32.
GaAs/GaNAs, SPF GaAs Luminescence, Raw Data,
Luminescent Spot for Probe Current 1x10-7 A ....................................37
GaAs/GaNAs, SPF GaAs Luminescence, Spline Data,
Luminescent Spot for Probe Current 1x10-7 A ....................................37
GaAs/GaNAs, Sample n554a, SPF GaAs Luminescence,
Spotwidth vs Spatial Samples..............................................................38
GaAs/GaNAs, Sample n546, SPF GaAs Luminescence,
Spotwidth vs Spatial Samples..............................................................38
GaAs/GaNAs, Sample n553, SPF GaAs Luminescence,
Spotwidth vs Spatial Samples..............................................................39
Computer Modeled and Measured Charge Motion Under
Applied Bias, Subtracted .....................................................................42
Computer Modeled Spot for Mobility Lifetime
Products 1.6x10-6, 1.8x10-6, and 3.48x10-6 cm2/V...............................42
x
ACKNOWLEDGMENTS
I would like to thank Professor Nancy Haegel for her help and for creating a great
academic environment within the laboratory to work in. The work presented in this paper
stemmed from many of her suggestions and leveraged off of her previous original work.
I would also like to thank Ben Cardozo at Lawrence Berkeley National
Laboratory for the GaAs samples and Homan Yuen at Stanford University for the QW
samples.
xi
THIS PAGE INTENTIONALLY LEFT BLANK
xii
I.
A.
INTRODUCTION
TRANSPORT IN SEMICONDUCTORS
Semiconductor technology is important in the development of advanced solid
state devices for electronics and optoelectronics. Semiconductors are used in ultraviolet
(UV) semiconductor lasers, solar cells, infrared (IR) detectors, and many other electronic
devices. Many IR and radar detectors/devices use semiconductor technology.
Semiconductors are unique in their ability to detect and produce photons across a wide
range of the electromagnetic spectrum.
Imaging charge carrier motion in semiconductors allows localized spatial
variation to be examined. The technique used in this analysis, charge transport imaging,
was introduced in 2003 and combines the spatial resolution of a scanning electron
microscope (SEM) with the optical imaging provided by a charge coupled display (CCD)
camera [1]. This technique allows monitoring of charge transport in luminescent
semiconductor materials while maintaining spatial imaging of the recombination process.
Keeping the charge generation at a fixed location (point), observation of diffusion and
drift by imaging the luminescence from the recombination is possible. The key
differences between this and cathodoluminescence (CL) is that the spatial resolution from
the recombination is maintained. This has the advantage for cases in which the diffusion
or drift of charge produces luminescence at locations away from the generation point.
In this work, drift behavior has been directly imaged in high purity epitaxial GaAs
under an applied bias. The technique has also been used to image the diffusive motion of
charge in epitaxial layers of n-type GaAs and other semiconductor materials.
Experimental data and simulated computer model generated results are compared.
Feasibility into using such techniques for contact-free measurements is also discussed.
B.
PURPOSE OF THIS THESIS
The purpose of this paper is to examine drift and diffusion in luminescent
semiconductors. This will be done for bulk materials which are characterized as three
dimensional (3D) and for thin epitaxial materials that can be approximated as two
1
dimensional (2D). These measurements were performed using a new experimental
technique called direct transport imaging. This was done utilizing a SEM coupled with a
thermoelectrically cooled CCD camera in the Physics Department at the Naval
Postgraduate School (NPS). A series of experiments was performed to measure
fundamental transport parameters in semiconductor materials. The goal was to extract
fundamental transport parameters with high spatial resolution. The feasibility and
limitations of this method for contact-free measurements are assessed.
C.
MILITARY RELEVANCE
Transport parameters are important to the development of new optoelectronic
materials and devices. Devices such as UV semiconductor lasers and advanced solar cells
for example are just some of the devices that could benefit from such a study. The ability
to image charge transport could potentially lead to being able to measure the effects of
defects on charge transport in semiconductor devices. Potential cost savings could also be
achieved by enabling transport characterization with minimal or no additional contact
processing. This could speed the development of new materials and reduce technology
transfer time for new semiconductor technologies.
D.
THESIS OVERVIEW
Chapter I describes an overview of the direct transport imaging technique and
describes the importance to military semiconductor devices. The need to image transport
in luminescent semiconductor devices leads to the investigation of diffusion and drift in
semiconductors, the main objective of this thesis. Fundamental concepts of charge
transport parameters are described in Chapter II. Chapter III discusses imaging transport
with and without an applied bias and comparisons are made to computer modeling.
Conclusions are discussed in Chapter IV.
2
II.
A.
TRANSPORT PARAMETERS
DIFFUSION AND DRIFT
Semiconductor materials are characterized by a bandgap of forbidden states,
separating a filled valence band (VB) from an empty conduction band of available
electron states. At temperatures above absolute zero, electrons are thermally excited from
the VB and generate electron-hole pairs. The average intrinsic carrier concentration (ni) is
constant for a given temperature. In parallel, some recombination will bring the electrons
from the conduction band (CB) to the VB. In direct bandgap semiconductors,
recombination yields emission of a photon. In indirect bandgap semiconductors, a
majority of recombination takes place through intermediate states by means of phonon
emission.
If this equilibrium is disturbed by an external excitation source, such as an
incident electron beam or from incident photons, additional carriers will be created. Let
nno and pno denote the electron and hole concentrations in the absence of excitation and nn
and pn denote the electron and hole concentrations with excitation (concentrations are per
volume). If the sample is n-type, the majority carriers are the electrons and the minority
carriers are the holes. Likewise, if the sample is p-type, the majority carriers are the holes
and the minority carriers are the electrons. For illustrative purposes, consider the n-type
case. Without excitation, it is known
nno pno = ni2
(1)
When excitation is present, equal number of electron-hole pairs will be created and of
course the product np will not equal ni2. However, the following two differences will be
equal [2]
∆nn = nn - nno = pn - pno = ∆pn
(2)
Since the material being considered is n-type, nno ~ Nd, and pno is found to be
much less from Equation 1. If the ∆nn changes by a small percentage, the ∆pn will of
course change drastically. Therefore, the minority carriers change drastically while the
majority carriers do not change much (percentage wise). Thus, the electrical properties
under excitation are often determined by the minority carriers.
3
Diffusion is the motion of carriers due to a concentration gradient, for example
formed by photogeneration or other excitation. Drift on the other had, is the motion of
carriers due to an applied bias. Currents in semiconductor devices are the result of both
diffusion and drift currents.
The diffusion length (Ldiffusion), which is the length a carrier will travel before it
recombines, can be found from the well known equation
L diffusion = Dτ
(3)
where τ is the mean free time (or lifetime as used in this context) and D is the diffusion
coefficient. The diffusion coefficient is found from the Einstein relation
D = kTµ/e
(4)
where k is Boltzmann’s constant, T is the temperature, and µ is the mobility. There is a
diffusion coefficient for electrons and holes. They differ in part by the mobility (µ =
eτ /m*) for the electrons and holes respectively (µe, µh), where m* is the effective mass of
either the electron or hole [3]. The constant e is the usual electron charge.
The drift length (Ldrift), which is the average length a carrier will travel due to an
applied electric field (E ), can be expressed as
Ldrift = µΕτ
(5)
The mobility (µ) appearing in the above equations is the constant of proportionality
between the drift velocity (vdrift) and the applied electric field. This can be expressed
mathematically simply as
vdrift = µE
(6)
In general, the total current in a semiconductor material can be expressed as J =
Jdrift + Jdiffusion + Jdisplacement (for both electrons and holes). At steady state, the
displacement term is zero and the current due to the electrons and holes is
Je = enµeE + eDedn/dx
(7)
Jh = enµhE + eDhdp/dx
The minority carrier lifetime is generally measured by what is known as the
photoconduction effect. The equation for this effect is [4]
Jpc = e(µe + µh)∆nE
(8)
4
where Jpc is the incremental current density as a result of the illumination, ∆n is the
number of electron-hole pairs per volume created by the illumination and the other
symbols have the usual meaning. Since ∆n = τ G resulting from photon generation rate
(G), the lifetime is given by [4]
τ = Jpc/[GEe(µe + µh)] ~ Jpc
(9)
This technique, however, requires precise knowledge of the generation rate, which can be
a complex combination of optical source, sample surface, absorption, and system optics.
Another standard technique of measuring minority carrier lifetime is using the
Stevenson-Keyes method [5]. The experimental set-up consists of a bias applied to a
sample with contacts and illuminated by a light pulse. An oscilloscope is connected at
one end which represents the load resistance. Solution to the differential equation for the
transport with appropriate boundary conditions is [4]
pn(t) = pno + τ hG exp(-t/τ h)
(10)
The decay is measured by the oscilloscope and this allows for determination of the
lifetime. Time resolved luminescence can also measure minority carrier lifetime. This
requires pulsed lasers and fast detectors that can match the wide range (approximately
10-3 to 10-12 sec) of observed lifetimes in semiconductor materials. This is an established
and widely used approach. However, it does not provide the complementary information
on mobility that is needed for diffusion or drift lengths, which depend on the µτ product.
One standard technique of measuring minority carrier drift mobility is that of
Haynes-Shockley [6]. Basically, localized light pulses are used to generate carriers in the
sample. Solution to the transport equation with appropriate boundary conditions with an
applied electric field yields [4]
p n (x, t) =
N
exp[-(x-µhΕt)2/(4Dht) - t/τ h] + pno
4π Dhτ
(11)
Again, an oscilloscope is used. By knowing the sample length, applied electric field, and
the time delay from the applied field and detected pulse (from the oscilloscope), the drift
mobility µ = x/(Εt) can be determined.
A disadvantage of many of the standard methods described is that the sample
must have contacts. Spatial uniformity is also assumed over the length of the sample.
5
This thesis will discuss a potential alternative approach that could lead the way for a
technique that does not require contacts on the sample and can provide spatial resolution
on the micron scale.
B.
LUMINESCENCE IN SEMICONDUCTORS
Luminescence is the emission of light, in excess of thermal radiation, caused by
external source excitation. There are several types of external excitation that can cause
luminescence. In photon-excitation, the source is in the form of photons (electromagnetic
radiation). An example of photo-excitation would be the photoluminescence (PL)
technique. Another type of external excitation is electron beam excitation in which an
energetic electron beam is used to bombard the sample to create luminescence. This is the
type of source excitation that is used in the direct transport imaging technique addressed
in this work. It is also commonly used in more conventional techniques such as CL.
There are other types of excitation, such as chemiluminescence and electroluminescence,
which are not pertinent to the discussion here [7].
When the incident electron or photon energy is greater than the bandgap, this
often leads in semiconductor materials to electron-hole pair production. When these
carriers recombine, light is produced at wavelengths corresponding to the allowed energy
states of the material (intrinsic, extrinsic) (∆E = 1.24/λ, where energy E is in eV and the
wavelength is in µm). This process is most efficient in direct bandgap materials.
In the case of electron beam excitation, many electron-hole pairs are created
through different mechanisms such as backscatter electrons, primary and secondary
electrons. Electron beam excitation is capable of producing many orders of magnitude
greater carrier generation than photo-excitation. This is one of the advantages to electron
beam excitation that can be useful in wide bandgap materials. There are two types of
scattering mechanisms, elastic and inelastic. Rutherford scattering of electrons by the
nuclei of atoms is an example of elastic scattering. Elastic scattering of electrons gives
rise to high energy backscattered electrons in the SEM [7]. Inelastic scattering may
include secondary electron emission, electron-hole pair generation, x-rays, and thermal
6
effects to name a few. Inelastic interaction processes are useful in electron probe
instruments.
When an electron beam is incident on a material, the electrons undergo a series of
successive elastic and inelastic scattering events. One result of this is the original
trajectories of the electrons are altered in a random fashion. Monte Carlo trajectory
simulation is one method of predicting the electron beam energy dissipation in solids, i.e.,
the generation volume and penetration depth. It is difficult to model the scattering path
though in a solid [7].
There are several statistical model expressions to estimate the electron penetration
depth. One general expression derived by Kanaya and Okayama has been found to agree
well with experimental results over a wide range of atomic numbers [8]. The penetration
depth estimate in µm is
Re = [0.0276A/(ρZ0.889)]Eb1.67
(12)
where Eb is in keV, A is the atomic molecular mass (g/mol), ρ is in g/cm3, and Z is the
atomic number. As an example, consider the case of bulk GaAs. At 35 keV electron beam
energy, the estimated penetration depth is approximately 6 µm. At 20 keV the penetration
depth is slightly over 2 µm while at 10 keV it is estimated to be slightly below 1 µm.
Other models such as the Everhart-Hoff expression [9] estimate somewhat lower
penetration depths. Nevertheless, all these expressions are useful in light of the fact they
give approximate penetration depths.
Electron beam excitation has the advantage of being able to control to a certain
degree the penetration depth into the sample material by varying the accelerating voltage
of the electron beam. In comparison to PL, where most of the incident photons are
absorbed very close to the surface and thus limit the depth of light penetration and
excitation volume, electron beam excitation can achieve much greater depth penetration
as discussed from the statistical simulation findings. As an example of PL penetration
depths, consider green light (λ = 514 nm) incident on GaAs. Most photons are absorbed
within the first 0.77 µm of the surface [10]. It is further noted that typically thousands of
electron-hole pairs are generated by a 20 keV electron while one photon generates only
one electron-hole pair in PL.
7
The carrier generation distribution, or just generation volume, is naturally of
interest in the study of electron beam excitation. A number of approximations for the
generation distribution have been used such as point source, sphere, and Gaussian. The
Gaussian models of the type
F(z) = Fo exp[ -a2(z - zo)2 ]
(13)
have been shown to provide accurate estimates [7]. Comparison to Monte Carlo
calculations [7] show that the generation volume is not perfectly defined. However,
contours of equal energy dissipation are in good agreement and indicate that most of the
energy is dissipated in a small volume near the electron beam impact point. The excess
minority carrier density is different for the one dimensional (1D), 2D and three 3D cases.
In the 1D case the excess minority carrier density decreases as ~ exp(-|x|/Ldiffusion) [11]. In
the 2D case the density decreases as ~ Ko(|x|/Ldiffusion) [1], where Ko is a zero order
modified Bessel function of the second kind, while in the 3D case the density decreases
as ~ 1/|x| exp(-|x|/Ldiffusion) [11]. Shown in Figure 1 is how the normalized excess minority
carrier density changes as a function of distance for the 1D, 2D, and 3D cases assuming a
point generation (with diffusion lengths of 1, 10, and 100 µm). It is readily seen that the
excess minority carrier density for the 3D case drops off rapidly. However, for the 2D
and 1D cases the excess minority carrier density does not drop off as rapidly.
This is important in transport imaging because these are parameters that affect the
imaged spotwidth. This allows understanding of the relation between carrier generation
and the region of luminescent recombination. We consider the effect of diffusion from a
point. The 3D case spot diameter is seen to be basically independent of Ldiffusion and
therefore primarily dependent on the generation volume. Donolato has shown that this
fact is what allows the resolution of CL defect imaging to be unaffected by carrier
diffusion [11]. However, the 2D case spot diameter is seen to reflect the carrier diffusion
length for diffusion lengths greater than the generation volume.
It should be expected for bulk materials that are definitely 3D, variation in the
spotwidth should not change much as a function of diffusion length. For thin sample
materials that can be approximated as 2D (or samples that can be approximated as 1D),
8
the spotwidth should change as a function of diffusion length. One experimental goal of
this work is to confirm this prediction.
9
(a) 1D.
(b) 2D.
(c) 3D.
FIGURE 1. Excess Minority Carrier Distribution as a
Function of Distance from Generation Point.
10
III.
A.
IMAGING TRANSPORT
DIRECT TRANSPORT IMAGING
Direct transport imaging is a relatively new technique that combines the charge
generation capability of a SEM, the resolution of an optical microscope, and the
sensitivity of a CCD camera [1]. The CCD is used to capture the luminescence of a
sample under electron beam excitation. This technique preserves the spatial information
of the charge recombination. This is the primary difference of the technique compared to
conventional CL. The electron beam in the SEM is used in spot mode to generate free
carriers in the sample. When the carriers recombine, the luminescence is captured by the
CCD camera.
The current system utilizes a JEOL 840A SEM. An optical microscope is inserted
into the chamber and connects externally to the thermoelectrically cooled Apogee CCD
camera. Shown in Figure 2 is the SEM and CCD camera experimental set-up. The silicon
detector in the CCD camera covers the band from approximately 400 to 1,100 nm. Any
unfiltered image would include all wavelengths within this band. The resolution of this
system is approximately 0.4 µm/pixel. The CCD is cooled to –15o C to minimize thermal
noise. The system consists of a mirror with a hole for the electron beam to pass through.
An optical microscope at the end of the apparatus magnifies and focuses the
luminescence from the sample onto the focal plane of the CCD array.
FIGURE 2. SEM and CCD Camera.
11
The SEM can collect data in either picture (pic) mode or spot mode (as well as
other modes not pertinent to the discussion here). In pic mode, the electron beam is raster
scanned across the sample. The luminescence is recorded by the CCD. This mode differs
from CL in that the camera exposure time can be much longer than a single frame in CL.
Traditional CL imagery uses a continuously scanned beam. In spot mode, the electron
beam of the SEM is held at a fixed location (point). The generation point can be precisely
controlled which allows for the local charge diffusion to be recorded. This can be done
with or without an external bias to study the mobility and drift of the minority carriers.
As an example, consider the case of bulk n-type GaAs doped at n = 5.9x1017 cm-3.
Shown in Figure 3 is the resulting image from the CCD in both pic mode and spot mode
for this sample. No applied bias is introduced. The probe current was set at 6x10-10 A, the
accelerating voltage was 35 kV, and the shutter exposure on the CCD camera was set at 1
sec. The images are 512 by 512 pixels, and since each pixel corresponds to approximately
0.4 µm at the sample, the resulting image size is approximately 205 by 205 µm.
400
0
50
362
100
6000
0
50
4562
100
325
150
200
288
250
3125
150
200
1688
250
250
250
300
300
350
350
400
400
450
450
512
512
0
100
200
300
400
512
0
(a) Pic mode.
100
200
300
400
512
(b) Spot mode.
FIGURE 3. Images of Bulk n-Type GaAs.
In Figure 3a, the area with the brightest intensity is the region over which the electron
beam was scanned. From the figure it is apparent at the area scanned in pic mode by the
SEM at this magnification (2000 X) is less than the area that which captured by the CCD.
The high intensity strip on the lower edge is a result of the SEM always scanning the
12
beam longer on one of the raster scan edges. In Figure 3b the light generated by the
electron beam at a fixed point is seen. It should be noted that the intensity scales on the
two figures are not the same, with the spot mode picture range being higher because of
the increased intensity focused at one location (point).
B.
DRIFT MEASUREMENTS
Using the direct transport imaging technique, measurements were made while
applying an external bias in order to image the drift motion of the minority carriers. The
sample (329SI) was n-type GaAs obtained from Lawrence Berkeley National Laboratory.
This sample was a high purity epitaxial layer doped at ND-NA = 5.1x1013 cm-3. The layer
thickness was approximately 60 to 80 µm on top of a semi-insulating GaAs substrate.
The sample had ion implanted contacts. The Au/Ge contacting metal layers were
evaporated on the top of the sample spaced approximately 2 mm apart. Shown in Figure 4
is a schematic top view of the sample and contact geometry.
Implants
~1 mm
Metal Contact
Metal Contact
~2 mm
FIGURE 4. Sample 329SI Contacts.
The implanted contacts spacing, though not visible, was approximately 1 mm. The SEM
accelerating voltage was set at 35 kV and the probe current at 3x10-8 A. The bias was
varied from +100 to –100 V. Thus, the highest applied electric field was approximately
100 V/mm. Shown in Figure 5 are images of the recombination distribution as a function
of applied bias. Notice that in the zero bias image (0 V) the spot is present without any
drift as expected. The drift of the charge is readily seen from the images captured by the
CCD. Since the material is n-type, the motion of holes is being imaged, i.e., the holes are
13
the minority carriers which are recombining to produce the illumination. The images are
512 by 1024 pixels which corresponds to approximately 205 by 410 µm. Comparison of
the 100 V image to the 0 V image reveals a long extended tail approximately 120 µm in
the figure imaging the charge carrier motion. For intermediate applied voltages between
+/-100 V and 0 V, the tail becomes less extended and broadens as the applied voltage
decreases as expected. The small circular spot next to the electron beam generated spot,
most noticeable in the zero bias case, is believed to be a result from a small reflection
within the system.
In the measurements discussed and shown in Figure 5, all the drift images include
the direct spot from the electron beam. In order to isolate and image the field induced
charge motion, the direct spot at zero bias was subtracted from the images taken at the
other biases. Shown in Figure 6 are images of the recombination distribution as a
function of applied bias with the direct spot subtracted. These are the same images as in
Figure 5 except that now the images have been subtracted. Notice in the 0 V image only
the mostly white background, corresponding to the noise background, is present. This is
as expected since the direct spot has been subtracted.
Other samples were measured besides the one shown in Figures 5 and 6. All
experimentally recorded measurements to date have shown a similar trend as illustrated
here for luminescent semiconductor materials. The results shown are characteristic of
such materials. Comparison of different materials using this approach could be used to
measure drift length and the material parameters (µτ) that determine it. Also,
measurement of the local field can check for field uniformity and/or voltage drop at
contacts.
14
100 V
90 V
80 V
70 V
60 V
50 V
40 V
30 V
20 V
10V
0V
FIGURE 5. Charge Motion Under Applied Bias.
15
-10V
-20V
-30 V
-40 V
-50 V
-60 V
-70 V
-80 V
-90 V
-100 V
FIGURE 5. (Contd.)
16
100 V
90 V
80 V
70 V
60 V
50 V
40 V
30 V
20 V
10 V
0V
FIGURE 6. Charge Motion Under Applied Bias, Subtracted.
17
-10 V
-20V
-30 V
-40 V
-50 V
-60 V
-70 V
-80 V
-90 V
-100 V
FIGURE 6. (Contd.)
18
C.
DIFFUSION MEASUREMENTS
1.
Bulk GaAs
Using the transport imaging technique, measurements were made on bulk n-type
GaAs sample doped at n = 5.9x1017 cm-3 without applying any external bias in order to
image the spotwidth under conditions of purely diffusive transport. The first set of
measurements made was the spotwidth as a function of electron accelerating voltage.
Since the differences in spotwidth between incremental accelerating voltages was on the
order of the resolution of the current optics in the system (approximately 0.4 µm/pixel), a
spline interpolation was used [12]. Shown in Figure 7 is the raw data and the spline
interpolated data at 30 kV. It is seen that the spline accurately interpolates the data
without distorting the raw data points. Spline interpolation was used on all data that
follows. The accelerating voltage was varied from 15 to 35 kV. Shown in Figure 8 are the
spots for various accelerating voltages. Since the background pixels are nonzero, the
spots are normalized so that the calculated half power spotwidths (referred to simply as
spotwidth in this paper) will be correct. The 2D images were recorded and 1D cuts
through the spots are shown in the figure. The x axis of the graph is displayed in units of
pixels with the center pixel being set to zero for ease of seeing the spotwidths. Figure 9
shows the measured spotwidth as a function of accelerating voltage. It is noted that the
spotwidth increases as a function of accelerating voltage. This is as expected since the
accelerating voltage is increasing the penetration depth of the electron beam and thus
increasing the generation volume of the minority free carriers. As discussed in a previous
section, we expect the generation volume to determine the luminescence spotwidth in the
thick sample limit.
Next the spotwidth was measured as a function of probe current. The accelerating
voltage was kept constant at 30 kV. Shown in Figure 10 are the 1D cuts of the spot for
probe currents varied from 1x10-10 to 1x10-7 A. Shown in Figure 11 is the spotwidth as a
function of probe current. It is seen from these graphs that the spotwidth does not change
and is basically constant in the bulk sample when the probe current is varied. The average
or mean was found to be 10.11 pixels with a standard deviation of 0.17 pixels. This is in
agreement with Donolato’s results. Since the sample is definitely 3D, we expect that the
19
generation volume should not change much as the probe current changes because the
generated spot size is less than the excitation volume diameter. For example, a probe
current of 6x10-12 A has a beam radius of approximately 5 nm, which is much less than
the generation volume radius. This corresponds to an area of 7.85x10-17 m2. As the probe
current is increased to 1x10-7 A, the area is increased to approximately 1.3x10-12 m2
which corresponds to a radius of approximately 645.5 nm or 0.646 µm (or perhaps
somewhat greater). This is still less than the penetration depth and generation radius at 30
kV.
Because of the large range of probe currents and accelerating voltages used in
these measurements, it is often desirable to vary the shutter time. Measurements of the
spotwidth as a function of shutter time on the CCD were made. The accelerating voltage
was held constant at 30 kV, the probe current was held constant at 3x10-9 A, and the
exposure time was varied. It was found that changing the exposure time did not affect the
normalized spotwidth measurements as long as the pixels were not saturated and it was
not noisy. If the exposure time was increased so that the center of the spot saturated the
pixels, the data of course was not useable. Thus, as long as the CCD is not saturated,
varying the exposure time did not affect the spotwidth measurements.
FIGURE 7. Bulk GaAs,
Luminescent Spot, Raw Data and Spline Interpolated Data.
20
FIGURE 8. Bulk GaAs,
Luminescent Spot for Accelerating Voltages 15 to 35 kV.
FIGURE 9. Bulk GaAs,
Spotwidth as a Function of Accelerating Voltage.
21
FIGURE 10. Bulk GaAs,
Luminescent Spot for Probe Currents 1x10-10 to 1x10-7 A.
FIGURE 11. Bulk GaAs,
Spotwidth as a Function of Probe Current.
22
2.
Epitaxial AlGaAs
Measurements were made on a 1 µm n-type AlGaAs epitaxial sample without
applying any external bias in order to image the spotwidth under conditions of diffusive
transport on a thin sample. This sample differs from the previous sample in that it can be
approximated as 2D. It was grown on semi-insulating GaAs which has weak
luminescence. Figure 12 illustrates the difference in excitation volume for a 2D and 3D
sample. The figure illustrates that diffusion length should be able to be measured without
luminescence from the area below. The first set of measurements made was the spotwidth
as a function of accelerating voltage. Measurements were taken with the probe current at
1x10-8 A. It was found that the spotwidth increases as the accelerating voltage increases,
as was expected. For example, at 10 kV the spotwidth was 3.32 pixels while at 20 kV the
spotwidth was found to be 3.94 pixels. This increase in spotwidth as a function of
accelerating voltage has been measured in a number of thin and thick samples.
Next, measurements were taken of the spotwidth while varying the probe current.
The accelerating voltage was held constant at 20 kV. Shown in Figure 13 are the spots for
various probe currents. It can be seen that as the probe current increases the spotwidth
increases. Shown in Figure 14 is a graph of the spotwidth as a function of probe current.
This increase in spotwidth is most likely due to being approximately in the 2D limit, i.e.,
the layer is thinner than the excitation volume. The increase in the electron beam from the
probe current increase is small. This series of experiments was repeated with an
accelerating voltage of 10 kV and the same trend was observed. The spotwidths as a
function of probe current are compared for 1µm AlGaAs and Bulk GaAs in Figure 15. It
is seen from this figure that the spotwidth changes only for the thin approximately 2D
sample.
23
Electron
Beam
Thin Sample
Thick Sample
FIGURE 12. Excitation Volume, 2D and 3D Sample.
24
FIGURE 13. 1 µm AlGaAs,
Luminescent Spot for Probe Currents 1x10-10 to 1x10-7 A.
FIGURE 14. 1 µm AlGaAs,
Spotwidth as a Function of Probe Current.
25
(a) Raw data.
4.4
Spot diameter (microns)
4.2
4.0
3.8
3.6
3.4
3.2
3.0
2.8
2.6
2.4
10-11
10-10
10-9
10-8
10-7
Probe current (A)
(b) Fitted data.
FIGURE 15. 1 µm AlGaAs and Bulk GaAs,
Spotwidth as a Function of Probe Current
(■ 1 µm AlGaAs, ● Bulk GaAs).
26
10-6
3.
GaAs/GaNAs Quantum Well
Since the AlGaAs data showed that the spotwidth varied as a function of probe
current for thin samples in the 2D approximation, this suggests the spotwidth is a
function of diffusive parameters rather than generation conditions for 2D samples. Thus,
a series of samples were required in which only diffusion lengths would be expected to
vary, with all other parameters such as material layer thickness and growth techniques
held constant.
The next samples measured were GaAs/GaNAs quantum wells (QWs) obtained
from Stanford University. There were three samples measured and all three had the
following layered structure.
500 Å
200 Å
3,000 Å
GaAs
GaNAs
GaAs
The main difference between the samples was the substrate growth temperatures. This
was of interest to this experiment since often higher growth temperatures lead to longer
diffusion lengths and better quality material. Other growth conditions were the same. All
of the samples have been rapid thermal annealed at 820 oC for 60 sec. This was done to
improve optical quality. Listed below are the growth temperatures for the three samples.
n554a
Substrate Temperature = 475 oC
n546
Substrate Temperature = 525 oC
n553
Substrate Temperature = 575 oC
The first surface layer of 500 Å most likely does not luminesce much as generally the
first tenth of a micron is usually dominated by surface recombination. Likewise, the bulk
substrate underneath the first three layers should not luminesce much either since it is
semi-insulating GaAs. Thus, the two layers that should luminesce most are the 200 Å
GaNAs layer and the 3,000 Å GaAs buffer.
The bandgap of GaAs is 1.42 eV corresponding to a wavelength of 870 nm, while
for GaNAs PL measurements showed the peak at 1,070 nm wavelength corresponding to
a bandgap of approximately 1.16 eV (measurements taken by Stanford). Due to the
27
differences in the bandgaps of the layers, it is possible to use a filter so that only one of
the two would be detected by the CCD camera at a given time. Measurements were taken
panchromatic (without using a filter), with a 950 nm long pass filter (allowing light from
the GaNAs layer through, while filtering out light from the GaAs layer), and with a 950
nm short pass filter (allowing light from the GaAs layer through, while filtering out light
from the GaNAs layer).
a.
Panchromatic
Shown in Figures 16 and 17 are the spots measured from the three samples
with a probe current of 1x10-7 A. Since this was measured without a filter, light from both
the GaAs and GaNAs layers was detected. It is seen that samples n544a and n553 have
spotwidths approximately the same. The spotwidth of sample n546 is the largest. Figures
18, 19, and 20 show the three different samples spotwidths taken at three different
locations on each sample for two different probe currents (3x10-8 and 1x10-7 A) to check
for reproducibility. It is seen that the standard deviations are small indicating the
measurements from the three different locations are about the same. The data shows that
even when the same probe current is used, sample n546 has a spotwidth that is larger than
that of samples n554a and n553 indicating differences in diffusive behavior. The general
trend does not show correlation between larger spotwidth and higher growth temperatures
(possible longer diffusion lengths), but since this data did not use a filter the spotwidth
was a function of light from both the GaAs and GaNAs layers. In order to isolate
illumination from the GaAs and GaNAs layers, filters were used.
28
FIGURE 16. GaAs/GaNAs, Panchromatic, Raw Data,
Luminescent Spot for Probe Current 1x10-7 A.
FIGURE 17. GaAs/GaNAs, Panchromatic, Spline Data,
Luminescent Spot for Probe Current 1x10-7 A.
29
FIGURE 18. GaAs/GaNAs, Sample n554a, Panchromatic,
Spotwidth vs Spatial Samples.
FIGURE 19. GaAs/GaNAs, Sample n546, Panchromatic,
Spotwidth vs Spatial Samples.
30
FIGURE 20. GaAs/GaNAs, Sample n553, Panchromatic,
Spotwidth vs Spatial Samples.
31
b.
950 nm Long Pass Filter, GaNAs Layer Luminescence
In order to isolate illumination from the GaNAs layer, a 950 nm long pass
filer (LPF) was used. Shown in Figures 21 and 22 are the spots measured from the three
samples with a probe current of 1x10-7 A. Since this was measured with a 950 nm LPF,
only light from the GaNAs layer was detected. It is seen that the spotwidth of sample
n546 is the smallest, while the spotwidth of sample n554a is the next largest, and the
spotwidth of sample n553 is the largest of the three samples. Figures 23, 24, and 25 show
the three different samples spotwidths taken at three different locations on each sample
for two different probe currents (3x10-8 and 1x10-7 A). It is seen that the standard
deviations are relatively large indicating differences in the measurements from the three
different locations. The data shows that when the same probe current is used, differences
in the spotwidths were measured. The general trend does not show correlation between
larger spotwidth and higher growth temperatures (possible longer diffusion lengths).
However, since the GaNAs process is not believed to be as good as that for GaAs, these
results may not be that surprising in that higher growth temperatures may not lead to
longer diffusion lengths in the GaNAs layer and there may be more variation. This may
be seen from the relatively large deviations in the data. For example, sample n553
showed a standard deviation as high as 0.27 pixels for a probe current of 3x10-8 A. This
may suggest spatial non-uniformity, which has been previously reported for MBE growth
GaNAs.
32
FIGURE 21. GaAs/GaNAs, LPF GaNAs Luminescence, Raw Data,
Luminescent Spot for Probe Current 1x10-7 A.
FIGURE 22. GaAs/GaNAs, LPF GaNAs Luminescence, Spline Data,
Luminescent Spot for Probe Current 1x10-7 A.
33
FIGURE 23. GaAs/GaNAs, Sample n554a, LPF GaNAs Luminescence,
Spotwidth vs Spatial Samples.
FIGURE 24. GaAs/GaNAs, Sample n546, LPF GaNAs Luminescence,
Spotwidth vs Spatial Samples.
34
FIGURE 25. GaAs/GaNAs, Sample n553, LPF GaNAs Luminescence,
Spotwidth vs Spatial Samples.
c.
950 nm Short Pass Filter, GaAs Layer Luminescence
The next set of measurements was done with a 950 nm short pass filter
(SPF). This would allow the light from the GaAs layer through while filtering out the
light from the GaNAs layer. Shown in Figures 26 and 27 are the spots measured from the
three samples with a probe current of 1x10-7 A. It is seen that the spotwidth of sample
n554a is the smallest, while the spotwidth of sample n546 is the next largest, and the
spotwidth of sample n553 is the largest of the three samples. The process of growing the
GaAs layer is believed to be much better than for the GaNAs layer. Generally speaking,
the minority carrier lifetime and therefore the diffusion lengths of GaAs will increase
with increasing growth temperature. Thus, the trend is as expected since the spotwidth is
increasing as a function of growth temperature (and more than likely the diffusion length)
for the GaAs layer. This indicates that the spotwidth for a sample that can be
approximated as 2D may be a function of the diffusion length.
Figures 28, 29, and 30 show data taken at three different locations on each
sample for two different probe currents (3x10-8 and 1x10-7 A). It was found that for the
probe current of 3x10-8 A sample n544a had a mean spotwidth of 3.36 pixels with a
35
standard deviation of 0.03 pixels, and for a probe current of 1x10-7 A had a mean
spotwidth of 3.74 pixels with a standard deviation of 0.00 pixels. It was found that for the
probe current of 3x10-8 A sample n546 had a mean spotwidth of 3.41 pixels with a
standard deviation of 0.03 pixels, and for a probe current of 1x10-7 A had a mean
spotwidth of 3.86 pixels with a standard deviation of 0.16 pixels. It was found that for the
probe current of 3x10-8 A sample n553 had a mean spotwidth of 3.77 pixels with a
standard deviation of 0.11 pixels, and for a probe current of 1x10-7 A had a mean
spotwidth of 4.37 pixels with a standard deviation of 0.21 pixels.
It should be noted that there was some deviations in the data. There were
two locations (two samples) that did not show the spotwidth increasing as a function of
growth temperature, though the average did show an increase as a function of growth
temperature. And there was never a case where the trend was backwards, i.e., the
spotwidth always increased (or remained the same in two of the measurements) as a
function of growth temperature. This data does indicate that there is potential to extract
diffusion parameters from the spotwidth on 2D samples with a technique that is sensitive
to both small changes in diffusion length as well as spatial variations. In the next section,
a numerical model will be applied to relate the variations in spotwidth to a quantitative
measure of diffusion lengths.
36
FIGURE 26. GaAs/GaNAs, SPF GaAs Luminescence, Raw Data,
Luminescent Spot for Probe Current 1x10-7 A.
FIGURE 27. GaAs/GaNAs, SPF GaAs Luminescence, Spline Data,
Luminescent Spot for Probe Current 1x10-7 A.
37
FIGURE 28. GaAs/GaNAs, Sample n554a, SPF GaAs Luminescence,
Spotwidth vs Spatial Samples.
FIGURE 29. GaAs/GaNAs, Sample n546, SPF GaAs Luminescence,
Spotwidth vs Spatial Samples.
38
FIGURE 30. GaAs/GaNAs, Sample n553, SPF GaAs Luminescence,
Spotwidth vs Spatial Samples.
39
D.
COMPARISON TO COMPUTER MODEL
A computer model to simulate diffusion and drift of minority carriers and their
recombination was developed [1]. The model includes diffusion and a 1D applied electric
field for the bias. The minority carrier distribution is given by
2
n
ϖ= 2
2π ( Ldiffusion )2
y+
1
1
x+
2n
2n
∫ ∫
y−
1
1
x−
2n
2n
( Ldrift )ξ
e
2( Ldiffusion ) 2
2
 ( L ) 2 + 4( L
drift
diffusion )

K0
2( Ldiffusion ) 2



r  dξ dη


(14)
where Ko is a zero order modified Bessel function of the second kind and r = (ξ 2+η 2)1/2.
The model was run for the case with a bias of 550 V/cm and Ldiffusion = 4.3 µm. As with
the earlier experimentally measured data, the final image was produced by subtraction of
images with and without the applied electric field. Shown in Figure 31a is the computer
model generated simulation of transport for this case [13]. The image size is 60 x 180
µm. As a comparison, shown in Figure 31b is the measured charge transport
recombination in epitaxial layers of n-type GaAs (329SI) with E = 550 V/cm. It can be
seen that the model produces results in close agreement with that measured.
The model was also run with the bias set to zero in order to model diffusive
behavior. The model currently exists in two forms, one that assumes a point source (Dirac
delta function) and another that allows for a finite generation region. The electron beam
incident on the surface of the sample is approximated to be square (rather than circular).
Setting the length of the sides of the square to 2 µm (which is a reasonable estimate),
spotwidths were calculated while varying the mobility lifetime product (µτ). Shown in
Figure 32 are the computer modeled spots for various mobility lifetime products. The
mobility lifetime products and corresponding spotwidths found were
µτ (cm2/V)
1.60x10-6
1.80x10-6
3.48x10-6
Spotwidth (pixels)
3.74
3.87
4.34
These spotwidths are approximately the same as the averages calculated for the
GaAs/GaNAs samples with the SPF (such that luminescence is from the GaAs layer) and
a probe current of 1x10-7 A. Comparison of Figure 32 to Figure 27 shows that the
modeled spotwidths are very close to the measured spotwidths. It is noted that there are
40
some slight differences in the spot shape, including background and thermal noise. These
may be attributed to the square approximation and other approximations. Further
improvements to the modeled are planned. It is seen from Figure 27 that the difference in
spotwidth from samples n554a and n546 looks greater than that modeled, however, this is
because Figure 27 is showing just one of the sampled locations. The means (from the
three locations) shown in Figures 28, 29, and 30 are approximately the same as that
predicted by the model. If the mobility is assumed to be 400 cm2/(Vs), the corresponding
lifetimes for the three mobility lifetime products (in order of increasing product) would
be 4.0, 4.5, and 8.7 nsec respectively. Thus, the initial results suggest that the technique,
in its current form, may be sensitive to changes of a factor of approximately 0.5 nsec in
lifetime.
One potential use of the transport imaging technique would be to measure local
diffusion lengths. Since the model has been shown to be in close agreement with
measured results, it could possibly be used to match the spotwidth of a measured sample
and thus the mobility lifetime product for the sample could be estimated. The ideal case
would be to take measurements on samples with known diffusion lengths. This was
unavailable at the time these measurements were taken and is something that is planned
for future experimental measurements. The advantage to using this approach, as opposed
to conventional techniques discussed earlier, is that the sample does not have to have
contacts. Contact-less diffusion measurements could provide a cost savings since the
expense of contact processing on the sample would not be there. Likewise for the applied
bias case, diffusion parameters could also potentially be estimated by matching the drift
in the measured images to that predicted by computer modeling. Knowing the applied
electric field and the mobility lifetime product would yield an estimate for the drift
length.
41
(a) Computer model.
(b) Measured data.
FIGURE 31. Computer Modeled and Measured Charge
Motion Under Applied Bias, Subtracted.
FIGURE 32. Computer Modeled Spot for Mobility Lifetime
Products 1.6x10-6, 1.8x10-6, and 3.48x10-6 cm2/V.
42
IV.
CONCLUSIONS
Imaging charge carrier motion in a number of semiconductor samples has been
demonstrated using the charge transport imaging technique both with and without the
application of external bias. Measured results were compared for both the biased and
unbiased cases to that predicted by computer modeling and the simulated results were
very close to that measured. It was found that for the unbiased case, spotwidths did not
change as a function of probe current on bulk 3D samples. However, the spotwidths were
found to change for thin samples that could be approximated as 2D as the probe current
was varied. Measurements were made using a SPF on GaAs/GaNAs QW samples so that
most of the luminescence would be from the GaAs layer. The difference between the
samples was that the growth temperatures were different. It was found that the spotwidths
increased as the growth temperature increased. Since the lifetime often increases as
growth temperature increases, this indicated that the spotwidth may be a function of
diffusive parameters for 2D samples.
A close examination of the spotwidth (from 2D intensity graphs) reveals that the
spot is not perfectly circular. This possibly could be a result of the electron beam not
being perfectly circular (remembering the current system has sub-micron scale
resolution). Because of this, it may be advantageous to use the area of the spot rather than
1D cuts through the spot [10]. Another area of improvement would be to replace the
optics to get better resolution. This would allow more accurate estimates of the
spotwidth. This is something that is planned in the future.
While it has been shown that the computer modeling and actual measured data are
in close agreement, the feasibility of contact-less diffusion measurements does require
some additional practical considerations. As discussed earlier, the ideal case would be to
get samples with known mobility and lifetimes to be used as a standard. Such samples
may be attainable to the degree required from the solar cell community, as knowledge of
these parameters is critical for their design and optimization.
Another possible use of the transport imaging technique using an applied bias is
for detecting defects in samples. Current interest particularly from the solar cell
43
community is for doing just that. Since this technique allows for sub-micron scale spatial
resolution, defects could potentially be mapped to a high degree of accuracy. Another
potential application is for mapping electric fields present between contacts due to the
space charge layer. While electric fields have been modeled using different techniques
(some neglecting the space charge layer), this could provide a means of measuring the
fields with good resolution. Both of these topics remain active areas of ongoing research.
44
LIST OF REFERENCES
1.
Haegel, N.M., Fabbri, J.D., and Coleman, M.P., “Direct Transport Imaging in
Planar Structures,” Appl. Phys. Letters 84, 1329 (2004).
2.
Karunasiri, G., “Semiconductor Device Physics,” Course Notes, Naval
Postgraduate School, Monterey, CA, 2004.
3.
Kasap, S.O., Principles of Electronic Materials and Devices, 4th Edition,
McGraw-Hill, New York, NY, 2002.
4.
Sze, S. Physics of Semiconductor Devices, 2nd Edition, John Wiley & Sons, New
York, NY, 1981.
5.
Stevenson, D.J. and Keyes, R.J., “Measurement of Carrier Lifetime in
Germanium and Silicon,” J. Appl. Phys. 26, 190 (1955).
6.
Haynes, J.R., and Shockley, W., “The Mobility and Life of Injected Holes and
Electrons in Germanium,” Phy. Rev. 81, 835 (1951).
7.
Yacobi, B.G. and Holt, D.B., Cathodoluminescence Microscopy of Inorganic
Solids, Plenum Press, New York, NY, 1990.
8.
Kanaya, K., and Okayama, S., “Penetration and Energy-Loss Theory of Electrons
in Solid Targets,” J. Phys. D: Appl. Phys. 5, 43 (1972).
9.
Everhart, T.E., and Hoff, P.H., “Determination of Kilovolt Electron Energy
Dissipation vs Penetration Distance in Solid Materials,” J. Appl. Phys. 42, 5837 (1971).
10.
Hoang, V.D., “Charge Transport Study of InGaAs QWIPS,” Master’s Thesis,
Naval Postgraduate School, Monterey, CA, June 2004.
11.
Donolato, C., “On the Theory of SEM Charge-Collection Imaging of Localized
Defects in Semiconductors,” Optik 52, 19 (1978).
12.
LabVIEW Function and VI Reference Manual, National Instruments, Austin, TX,
1998.
13.
Haegel, N.M., Hoang, V.D., and Freeman, W., “Imaging Transport: Monitoring
the Motion of Charge Through the Detection of Light,” ICPS, July 2004.
45
THIS PAGE INTENTIONALLY LEFT BLANK
46
INITIAL DISTRIBUTION LIST
1.
Defense Technical Information Center
Fort Belvoir, VA
2.
Dudley Knox Library
Naval Postgraduate School
Monterey, CA
3.
Professor James H. Luscombe, Chairman, Department of Physics
Naval Postgraduate School
Monterey, CA
4.
Professor Nancy M. Haegel
Naval Postgraduate School
Monterey, CA
5.
Will Freeman
NAWCWD
China Lake, CA
47